IRS13 Search Matching Functions chapter E. M. Keen Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. 111-8 C) Automated Systems These illustrations are given to highlight some of the problems inherent in conventional systems, as an introduction to the [OCRerr] methods used with the SMART systeme The objectives of introducing such methods in a retrieval situation are: 1. To avoid the need for manual search logic formulation; 2. To minimize search time not by using logical relation demands but by asking only for request/document term matches using, if neces- sary, weights associated with each term; 3. To present to the user an ordered list of documents arranged in decreasing order with the search request correlation (ranked output) so that the cut-off may be a user decision at the output stage. Fully automated retrieval Systems are characterized by the replacement of human intellectual effort where it can be as efficiently performed by a machine, and also on occasion by the provision of a man-system interaction that permits the human1 5 irreplaceable contribution (decisions as to cut-off point, judgment of value of documents examined, etc.) to be entered as a control in the search process. The SMART system is investigating the design and evaluation of such automated systems; the use of algorithms that establish matching coefficients between search requests and documents is a part of such a task. One question to be answered is whether it is necessary to use any human judgment during the search or matching process. In addition, techniques of "weighting'1 search terms according to some criterion of importance appear to be worth investigating, since they may easily be incorporated into automated systems. The results from the SMART experiments give insight into matching functions for automated systems, as the following results show.